Date: Fri, 16 Feb 1996 16:55:19 -0500 From: "Louis A. Mamakos" <louie@TransSys.COM> To: dennis@etinc.com (dennis) Cc: hackers@FreeBSD.ORG, isp@FreeBSD.ORG Subject: Re: Frame Relay and FreeBSD Message-ID: <199602162155.QAA21088@wa3ymh.transsys.com> In-Reply-To: Your message of "Fri, 16 Feb 1996 10:19:14 EST." <199602161519.KAA02788@etinc.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> Louis writes..... > > >On the other hand, your Internet gateway doesn't have to fsck when the > >power fails and comes back again. It doesn't serve web pages, or run > >SMTP.. it does actually route packets pretty well. > > You're missing the point...you're comparing a router to a host/router, and > I'm comparing a card to a router box.. No....my card doesnt have > any moving parts either. There are substantial performance and functional > advantages to having a common code solution between router and server. > What's really funny is that you contradict yourself in your next "argument" No, you're missing the point. In my original message I pointed out that the cost of a dedicated Ascend Pipeline 50 LS56 is very close if not cheaper than the cost of the synchronous interface for a PC platform and the associated external CSU/DSU. The moving parts refer to the disk drive in the PC that loads the OS that makes the interface do anything useful. If you only have one machine, then the failure of the disk isn't any big deal. If, on the other hand, you have a local ethernet with a handful of machines on it, why would you tie their functioning to a realtively delicate platform with moving parts? > >To each his own.. My PL50 works really well, and takes up less space > >than the ZyXEL modem I used to use. It doesn't have moving parts, and > >doesn't need an external CSU/DSU. > > My argument is to put a router INTO my server...your argument of putting the > CSU into the router is similar, except that you almost always dont have a choice > of which CSU/DSU to use with an integrated solution. If you dont care...then > fine...but they're almost never as good or reliable as external ones. You only need a working CSU/DSU that conforms to the relevant Bellcore specs. I don't know why you'd believe the external CSU/DSUs are more reliable since they all use the same VLSI these days. > > > >> > >> A good rule of thumb: If you know un*x, use un*x....if you don't, use > >> something else. > >> > > > >Yes, it's true that if the only tool you have is a hammer, everything > >looks like a nail. That's no excuse not to get the 'right' tool for > >the job. > > I dont think that the "hammer" argument works here...theres no evidence > that your solution is better or more appropriate than a unix solution, and > much evidence that it is less flexible and lower performance (due to an > added hop, at the minimum). From the talk about problems with the PLs, > I'd say its a pretty risky choice. It absolutely is more reliable; the performance of your entire network doesn't depend on the functioning of a UNIX box with disk drives, etc. When there's a power failure, I don't want the connectivity to my network to fail until someone comes along and types 'fsck -y' on the console to reboot the gateway. Louis A. Mamakos
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199602162155.QAA21088>